Skillv1.0.0

ClawScan security

多 Agent 混合层级隔离架构 1.0 · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignMar 8, 2026, 2:51 AM

Verdict: Benign
Confidence: high
Model: gpt-5-mini
Summary: The skill's files, instructions, and checks are coherent with its stated purpose (a documentation + templates-based multi‑Agent isolation architecture); it does not request extraneous credentials, network downloads, or install steps and primarily provides templates and a local config-check script.
Guidance: This package is documentation + templates for separating a 'writer' and 'media' agent and a local config-check script — overall coherent. Before installing/applying templates: 1) Review the SOUL.md templates (writer/media) to ensure you are comfortable with the enforced constraints and wording (they change agent behavior). 2) Back up any existing SOUL.md and openclaw.json before overwriting or copying files. 3) The config-check.sh reads local OpenClaw files (~/Documents/openclaw/agents, ~/.openclaw/openclaw.json) and checks for the openclaw binary and logs; run it locally to validate environment but inspect its output and exit codes. 4) The recommended softlink approach shares skills between agents — verify that writer does not get unintended access to baoyu-* skills via your filesystem permissions or link locations. 5) No sensitive credentials are requested by this skill, but treat any template that changes agent behavior as granting behavioral control: only apply templates you trust. If you want lower risk, copy templates to a separate test environment first and validate behavior before applying in production.

Review Dimensions

Purpose & Capability: okName/description (multi-Agent hybrid isolation) match the delivered artifacts: docs, SOUL.md templates for writer/media, and a config-check script. The requested files and operations (copy templates, create softlinks, check openclaw.json, restart gateway) are consistent with building the described architecture.
Instruction Scope: okSKILL.md instructs only local configuration actions: copying template SOUL.md files, creating soft links for shared skills, running a local shell script, and restarting the local OpenClaw gateway. There are no instructions to read unrelated system files, phone home, or send data to external endpoints.
Install Mechanism: okThere is no remote install/download in the package. It's instruction-only with local templates and a shell script. No URLs, archive extraction, or third-party package pulls are embedded in the skill itself.
Credentials: okThe skill declares no required env vars or credentials. The script and docs reference local paths (~/Documents/openclaw/agents, ~/.openclaw/openclaw.json, /tmp/openclaw logs) which are appropriate for an OpenClaw agent configuration helper. No unrelated credentials or secret access is requested.
Persistence & Privilege: okSkill flags are default (always: false) and it does not attempt to modify other skills or system-wide configs beyond advising copying templates and restarting the gateway. The scope is limited to changing agent SOUL.md and recommending soft links for shared skills.